April, 19661 HINE, CRAWFORD, DEUTSCHMAN AND TIPTON 241 The Determination of Sodium in Aluminium Alloys Flame Spectrophotometry with Fuel-rich Flames to Reduce Interference BY R. A. HINE,* R. CRAWFORD,? J. E. DEUTSCHMAN AND P. J. TIPTON: (Aluminium Company of Canada, Limited, A rvida, Quebec, Canada) The determination of trace quantities of sodium in aluminium alloys by flame spectrophotometry offers a rapid and accurate control procedure. When conventional flames with a balanced fuel - air mixture are used, molecular oxide band spectra of iron and manganese are strongly excited, and interfere with the measurement of the sodium emission. The interference is much less in fuel-rich flames while the sensitivity to sodium is slightly increased. The use of fuel-rich flames therefore provides a more versatile and accurate method than those hitherto used. As in most methods for the determination of sodium it is advantageous to add lithium as an internal standard to compensate for minor variations in conditions. The results are of general application to the analysis of other materials which may contain iron and manganese, and possibly nickel and chromium.THE deleterious effects of small amounts of sodium in aluminium - magnesium alloys were first described by Ransley,l who showed that as little as 10 p.p.m. of sodium could cause embrittlement leading to cracking on hot-rolling of the ingots. With large ingots even smaller sodium contents have adverse effects, and it is therefore necessary to be able to carry out accurate analysis before fabrication of the ingot.Spectrographic determination of sodium is used in routine control but requires chemically- analysed metal standards for calibration. Careful attention to the sampling procedure is also necessary because of the tendency for sodium to segregate. Flame spectrophotometry offers a highly sensitive means of determining sodium,2 and has been used to establish the calibration values of spectrographic standards. Moreover, samples can also be taken by drilling the solid ingot, thus avoiding segregation difficulties. Matelli3 described a method using a Beckman DU spectrophotometer with flame attachment capable of determining down to 10 p.p.m. of sodium in aluminium alloys. By using the more sensitive Unicam SP900 flame spectrophotometer, Hine and Rates4 extended the range to 1 p.p.m.without using methanol to intensify the emission. However, in applying the latter method to the analysis of a variety of aluminium alloys it has been found that errors can result from interference unless the metal used for calibration is identical in composition (other than sodium content) with the sample being analysed. In particular, differences in iron and manganese contents between the sample and calibration metal give rise to serious errors. In this paper it is shown that the extent of interference is governed by the type of flame used and that with cool, fuel-rich flames the interference is slight. EXPERIMENTAL APPARATUS- Flame spectrophotomeler-~nicam SP900 with acetylene - air flame. All other apparatus should be made of quartz, translucent silica or polythene, as appropriate.REAGENTS- Sodium-free alloy for calibration-Place a clean piece (10 to 50 g) of alloy, of the same type as that to be analysed, in a graphite crucible in a vacuum distillation apparatus capable of maintaining a vacuum of at least 10-5 mm of mercury. A suitable apparatus is described * Present address : Aluminium Laboratories Limited, Banbury, Oxon. t Present address : International Alloys Limited, hylesbury, Bucks. Present address: Canada Colors & Chemicals Dominion Ltd., Toronto 3, Ontario.242 HINE, CRAWFORD, DEUTSCHMAN AXD TIPTON: DETERMINATION [Analyst, Vol. 91 on p. 24 of “Analysis of Aluminium and its Alloys.”2 Heat to 900” C for 1; hours. Cool, remove the specimen, pickle the surface in dilute hydrochloric acid, rinse and dry.Repeat the vacuum treatment, Reduce by milling or drilling with scrupulously clean tools. Take 1-g portions of the metal for calibration. Pure water-Pass water (preferably distilled) through a mixed-bed resin, e.g., Amberlite MB3, using all-plastic apparatus. Hydrochloric acid, 6 N-Prepare from the purest obtainable reagent ; if necessary, distil in quartz. Sodium chloride stock solution-Dry pure sodium chloride at 105” C ; dissolve 2.5418 g in water and make up to 1000 ml in a volumetric flask. Transfer immediately to a polythene bottle. Sodium chloride working solutions-Prepare dilutions of the stock solution such that 1 ml = 10 pg of sodium and 1 ml = 1 pg of sodium, respectively. Lithium chloride stock solution-Dissolve 5-32 g of lithium carbonate in a minimum of dilute hydrochloric acid in a platinum dish.Heat gently to expel carbon dioxide, cool and make up to 1000 ml in a volumetric flask. Transfer immediately to a polythene bottle. 1 ml = 1 mg of lithium. Lithium chZoride working solutions-Prepare dilutions of the stock solution such that 1 ml = 100 pg of lithium and 1 ml = 10 pg of lithium, respectively. Other metals used were of Johnson and Matthey “Specpure” grade. 1 ml = 1 mg of sodium. PROCEDURE (FOR RANGE 0 TO 10 P.P.M. O F SODIUM)- Weigh 1 g of sample into a 250-ml silica beaker; at the same time weigh six 1-g samples of sodium-free alloy for calibration into other beakers. Wash beakers and contents by decantation three times with de-ionised water, carefully draining off the water each time.Add 30ml of 6~ hydrochloric acid, cover with a silica watch-glass and warm to start the reaction, finally heating to complete solution. Evaporate to incipient crystallisation of salts and allow to cool. Re-dissolve the salts in 25 ml of water and transfer (filtering if necessary) into 100-ml quartz volumetric flasks. Add to each 5 ml of lithium solution (1 ml = 10 pg of lithium) and to five of the calibration solutions add 2, 4, 6, 8 and 10 ml of standard sodium solution (1 ml = 1 pg of sodium) retaining the remaining one as a blank. Adjust to volume with de-ionised water, mix and transfer to clean polythene bottles. Set up the flame photometer using suitable slit and gain control settings and adjust the flame as indicated below. Aspirate the highest standard and arrange to give 90 to 95 per cent.galvanometer deflection when set on the sodium line at 589 mp. Aspirate all the standards and samples, clearing between each by passing de-ionised water, and record the sodium emission readings. Re-set the instrument to the wavelength of the lithium peak at 670-7 mp to give a galvanometer deflection of 40 to 60 per cent. and again aspirate standards and samples, recording the lithium emissions. Subtract the blank sodium reading from the readings for the standards and samples. Plot the sodium-to-lithium emission ratios against sodium content and read the contents of the samples from the graph. For sodium contents above 10 p.p.m. use the stronger sodium and lithium solutions for calibration, with appropriately lower gain settings on the instrument.NOTE-In the case of alloys containing Inore than 1 per cent. of silicon, ignite the filtered silicon Dissolve residue in a platinum crucible, treat with hydrofluoric and nitric acids and evaporate to dryness. the residue in a little hydrochloric acid and add to the main solution. FLAME BACKGROUND- The air was set at 23 p.s.i. throughout, as lower pressures are insufficient to give good atomisation of the solution. Variation of flame conditions was obtained by adjusting the gas pressures. For the examina- tion of traces of sodium it is obviously necessary to keep contamination and over-all back- ground down to a minimum. The background at 589 mp was therefore measured at a series of acetylene pressures using a solution of sodium-free N8* alloy (1 g per 100 ml).The instrument settings were arranged to give full deflection of the galvanometer with a solution containing 0.1 pg of sodium per ml using pure water as the reference solution. An acetylene - air flame was used in these experiments. * The alloy designations used throughout are those of B.S. 1470.April, 19661 OF SODIUM IN ALUMINIUM ALLOYS BY FLAME SPECTROPHOTOMETRY 243 The results are shown in Fig. 1, from which it is seen that minimum background is obtained at positions A or B. At position B the flame is fuel-rich and just verging on luminosity. At position A it is air-rich and unstable. Position C represents the balanced flame usually used and recommended by the manufacturers; it is achieved by starting with a gas pressure of about 10 cm and turning down the gas until the blue burner cones just verge on instability.Surprisingly, this conventional flame gave high background, and the gas-rich flame type I3 appeared to be preferable. However, this experiment alone was not conclusive since background is composed of several factors, including sodium impurity in the reagents, effects of reagents on the flame and, possibly, effects of alloying elements in the metal. Conceivably, the higher emission with flame C could be reflecting a greater sensitivity to the sodium impurity unavoidably present. 24 $22 rn co In % 20 c .- v) .- E I* w 16 26 I - - - - - I I I 141 I I I 1 1 - 2 4 6 8 10 12 14 Acetylene pressure, cm O' I ; ; 1 ; Q ; pg of sodium per 100 ml Fig. 1. Effect of flame condition on back- Fig.2. Calibration graphs for different flame conditions. Acetylene pressure: (A) 8 cm; (B) 14 cm; (C) 4 cm ground emission a t 589 mp. A: Air-rich and un- stable flame; B: Fuel-rich flame and C: Balanced flame Calibration curves were therefore constructed for the three types of flame, again using N8 alloy as the calibration metal. The results are shown in Fig. 2. This figure demonstrates again the higher background with the conventional balanced flame, but also shows that the sensitivity to sodium is actually better in the other types of flame. To determine the reason for the higher background intercept with the conventional balanced flame, other experiments were carried out. No effect was found by adding magnesium up to the equivalent of 10 per cent.to calibration solutions, hence the magnesium content of S8 alloy was not the cause. This alloy also contains about 0-8 per cent. of manganese, 0-3 per cent. of iron and 0.2 per cent. of silicon (the silicon is removed in processing so this could not be the cause of the background). Similar behaviour in different flame types was found with commercially pure aluminium (0-4 per cent. of iron, 0.3 per cent. of silicon) but not with super-purity metal (99.89 per cent. of aluminium). Thus attention was given to the effects of iron and manganese on background emission. EFFECT OF IRON AND MANGANESE Both iron and manganese are shown by Dean5 to exhibit pronounced molecular oxide bands in the region of sodium emission. The exact situation at 589 mp is difficult to assess because of sodium impurity in the solutions used, but it seems possible that excitation of oxides could vary considerably in different flame types and hence give rise to the anomalous effects observed.Solutions were prepared containing 1 mg and 5 mg of iron per 100 ml, and also 5 and 10 mg of manganese per 100 ml. The spectra recorded showed much higher emission from iron and manganese in balanced and air-rich flames than in the fuel-rich flame. However, the presence of sodium impurity in these solutions complicates the interpretation, and a more convincing experiment was carried out. Duplicate samples of aluminium were dissolved in hydrochloric acid and 10 mg of iron added to the solution. Iron was extracted from one solution with di-isopropyl ether and the spectra of the solutions then recorded in244 HINE, CRAWFORD, DEUTSCHMAN AND TIPTON: DETERMINATION [Analyst, Vol.91 three types of flame. The results are shown in Fig. 3. In each case the broken line shows the emission from sodium only (after iron extraction) while the full line shows the spectrum with superimposed emission from molecular iron oxide. These results show conclusively that iron oxide band spectra interfere seriously in the determination of sodium when conventional balanced or air-rich flames are used for excitation, but that interference is slight in a fuel- rich flame. Wave I ength, mp Wavelength, rnp L I 580 600 t Wavelength, rnp Fig. 3. Spectra of solutions before and after extraction of iron (10 mg per 100 ml). Broken lines, iron-free spectrum; continous line, iron with sodium impurity.( a ) air pressure, 23 p.s.1. ; acetylene pressure, 14 cm; (b) air pressurc, 23 p.s.i., acetylene pressure, 8 cm; (c) air pressure, 23 p.s.i., acetylene pressure, 4 cm The effect of increasing iron contents on the emission from 5 pg of sodium per 100 ml, in the presence of aluminium, is shown in Fig. 4 for the three flame types. Similar results were obtained for manganese additions except that the magnitude was lower with manganese. Again it is obvious that the fuel-rich flame is least sensitive to interference and gives an almost horizontal line, whereas the most air-rich flame has the steepest slope. MAGNITUDE OF ERRORS The magnitude of errors caused by iron or manganese interference depends on the relative contents of these elements in the calibration metal and the sample.The errors will be positive if the sample contains more of these elements than the calibration metal and negative if it contains less. Thc crror is always at a minimum with a fuel-rich flame. Some examples are given in Table I. In this table the figure given as the established value is the best estimate of sodium content obtained by a radioactivation method. The calibration metal was commercially pure aluminium containing about 0-25 per cent. of iron. TABLE I RESULTS OBTAINED ON ALUMINIUM ALLOYS IN FUEL-RICH AND AIR-RICH FLAMES Composition, per cent. Established 7- sodium, iron manganese p.p.m. 1.2 0.04 2 0-22 0.3 7 0-3 0.8 1.1 0.16 0.17 3.2 0.01 0.0 1 < 0.5 I Fuel-rich flame sodium found, p.y .m. 3.3 6.3 1.5 3.8 0 11 Air-rich flame sodium found, 8-5 Serious iron effects in I1 10 Manganese effect in I1 4.0 Manganese effect in I1 5.0 - 2-2 Negative errors in I1 clue to lower iron content of sample p.p.m. Remarks -April, 19661 OF SODIUM IN ALUMINIUM ALLOYS BY FLAME SPECTROPHOTOMETRY 245 Iron, mg per 100 ml Fig. 4. Effect of increasing iron on emission of 5 pg of sodium per 100 ml. Acetylene pressure: Graph A (A), 8 cm; graph B (e), 14 cm; graph C (O), 4 cm FLAME REPRODUCIBILITY AND PRECISION USING FUEL-RICH FLAME The fuel-rich flame shown to be preferable from the point of view of interferences is less familiar than the conventional balanced flame, and possibly less convenient to work with. However, after some experience it is possible to reproduce the flame and obtain consistently good results.The general stability of the instrument under these conditions is excellent, as shown by the following experiment. A solution of an aluminium - magnesium alloy con- taining 3.5 p.p.m. of sodjum was prepared, the instrument set up and a calibration carried out; 6 readings were taken on the sample solution and the sodium content determined. The flame was then turned out, re-lit, re-adjusted and the procedure repeated to a total of 6 times. The “within-run” standard deviation found was equivalent to t-0-15 p.p.m. of sodium and the “between-run” value was also &0-15 p.p.m. The over-all precision of the method, established by processing separate weighed samples of several alloys was found to be about &O-5 p.p.m.for a content of 3 p.p.m. of sodium. DEFINITIOK OF FLAME TYPES The figures for fuel pressure given in the text, and illustrations as representing the flame types used, only provide a general indication of the value. The exact pressure to use depends upon the age and condition of the burner jet and other variables. Final adjustment of the flame is therefore carried out visually according- to the following instructions. (For the Unicam SP900 instrument, with air pressure at 23 p.s.i., the approximate acetylene pressures are shown in brackets.) Bnlmzcecl pame-This is the conventional flame in which there is sufficient air to give steady and complete combustion of the fuel. I t is obtained by starting with a slightly fuel-rich flame and then turning down the gas pressure until the inner blue burner cones just become unstable, then increasing the pressure slightly until the cones regain stability (8 to 10 cm of acetylene pressure).Air-richflame-The fuel pressure is turned down to the point where the flame will only just remain alight. The flame is weak, short and barely visible (4 to 5 cm of acetylene pressure). Fuel-richjame-The fuel pressure is turned up until the flame becomes luminous and then turned back until the luminosity just disappears (11 to 14 cm of acetylene pressure). DISCLJSSION AKD CONCLUSIONS The interference caused by excitation of molecular oxide bands of iron and manganese was the most important aspect of a larger investigation into the determination of trace amounts of sodium in aluminium alloys. Other work showed that the acidity of the solution246 HINE, CRAWFORD, DEUTSCHMAN AND TIPTON [Analyst, Vol.91 affected the emission from sodium and that there were complex interactions between various factors. However, by close standardisation of the procedure, and the use of lithium as an internal standard to compensate for small variations in conditions, accurate results could be obtained. The errors caused by iron or manganese, or both, when a conventional flame is used, are of sufficient magnitude that the calibration metal must be of almost identical composition with the sample to obtain correct results. In practice this would mean subjecting a portion of each sample to vacuum distillation to serve as its own calibration metal. On the other hand, although some interference still occurs in fuel-rich flames, it is so small that calibration metal of similar alloy type to the sample can be used without incurring appreciable error.In fact, several groups of alloys, such as the aluminium - magnesium series, are sufficiently similar in iron and manganese contents to allow common standards of sodium-free calibration metal to be used. Of other common alloying elements, it has been found that copper, zinc and magnesium have no appreciable effect, but a few less common elements such as nickel and chromium (which are known to give oxide bands) may have similar effects to iron and manganese. However, the use of calibration metal of the same alloy type in conjunction with a fuel-rich flame would be likely to give accurate results in these cases also.It is not clear whether the lower emission from iron and manganese oxides in fuel-rich flames is a result of the lower temperature of such flames, or whether the reducing character of the flame suppresses the formation of oxides. Fasse16y7 and his co-workers have carried out extensive investigations into the flame spectra of rare-earth elements and of vanadium, niobium, etc. They have shown that, whereas in conventional (stoicheiometric) flames monoxide band spectra of these elements predominate, in fuel-rich flames distinctive line spectra appear which allow quantitative determination of the elements. Fassel considers that though the fuel-rich flames are cooler, the main reason for the predominance of line spectra is that the conditions for monoxide formation are unfavourable (reducing species present) and lead to a high atom population in the flame.A similar explanation fits all of the observations made in the present work. Analogous effects were also found in the deter- mination of lithium in aluminium alloys, and the wider investigation of unconventional fuel-rich flames may result in improvements in flame spectrophotometry procedures for many elements. Although the work was concerned with the analysis of aluminium alloys, the results are of general application, and could be of significance in the determination of sodium in biological samples, plant materials and minerals. For all analyses where iron or manganese, or both, may be present, the use of a fuel-rich flame, just verging on luminosity will give the greatest freedom from interference coupled with high sensitivity. REFERENCES 1. 2. 3. 4. 5. 6. 7. Ransley, C. E., and Talbot, D. E. J., J . Inst. Metals, 1959-60, 88 (4), 150. “Analysis of Aluminium and its Alloys,” The British Aluminium Company, Limited, Publication No. 405. Mattelli, G., and Attini, E., “Determination of Sodium in Aluminium by Flame Spectrophoto- metry,” Instituto Sperimentale dei Metalli T.eggcri, Publication No. X-243, 1960. Hine, R. A., and Bates, J. F., “Applied Materials Research,” October, 1963, 216. Dean, J . .4., “Flame Photometry,” McGraw-Hill Book Company Inc., Sew Yorli, 1960. Fassel, V. A,, Curry, R. H., and Kniseley, R. N., Spectrochzm. A d a , 1962, 18, 1127. Fassel, V. A., RiZyerP, R. B., and Kniseley, R. N., Ibzd., 1963, 19, 1187. Received August lGtJz, 1965